Research map / Continual learning / Class-incremental
Incremental Learning Using Conditional Adversarial Networks
Ye Xiang, Ying Fu, Pan Ji, Hua Huang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add adversarial loss (ablation) | Ours · CIFAR-100 (incremental step 10) | alpha_all: 0.654 -> 0.671 | Improved |
| add adversarial loss (ablation) | Ours · CIFAR-100 (incremental step 10) | alpha_orig: 0.623 -> 0.644 | Improved |
| add reconstruction loss (ablation) | Ours · CIFAR-100 (incremental step 10) | alpha_all: 0.663 -> 0.671 | Improved |
| add reconstruction loss (ablation) | Ours · CIFAR-100 (incremental step 10) | alpha_orig: 0.635 -> 0.644 | Improved |
| add adversarial and reconstruction losses to classification loss (ablation) | Ours · CIFAR-100 (incremental step 10) | alpha_all: 0.628 -> 0.671 | Improved |
| add adversarial and reconstruction losses to classification loss (ablation) | Ours · CIFAR-100 (incremental step 10) | alpha_orig: 0.597 -> 0.644 | Improved |
| add adversarial and reconstruction losses to classification loss (ablation) | Ours · CIFAR-100 (incremental step 10) | alpha_new: 0.842 -> 0.830 | Hurt |
| remove adversarial loss effect on new-class accuracy (ablation) | Ours · CIFAR-100 (incremental step 10) | alpha_new: 0.838 -> 0.830 | No clear effect |
| store statistics (s-way) instead of directly storing embeddings (d-way) (ablation) | Ours · CIFAR-100 (step 10, t=5) | alpha_all: 0.590 -> 0.591 | No clear effect |
| fine-tune convolutional layers too (Ours vs Ours-R) | Ours · CIFAR-100 (incremental step 10) | alpha_all: 0.625 -> 0.671 | Improved |
| fine-tune convolutional layers too (Ours vs Ours-R) | Ours · CUB-200 (incremental step 10) | alpha_all: 0.520 -> 0.576 | Improved |
| use conditional adversarial generator replay instead of FearNet | Ours · CIFAR-100 (incremental step 25) | alpha_all: 0.663 -> 0.670 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- FearNet — FearNet: Brain-Inspired Model for Incremental Learning (ICLR 2018) · better on 6 of 6 shared results
- iCaRL — iCaRL: Incremental Classifier and Representation Learning (CVPR 2017) · better on 6 of 6 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Class-Incremental Learning: Survey and Performance Evaluation on Image Classification (TPAMI 2022 · similarity 0.90)
- Incremental Learning of Object Detectors Without Catastrophic Forgetting (ICCV 2017 · similarity 0.89)
- IL2M: Class Incremental Learning With Dual Memory (ICCV 2019 · similarity 0.88)
- Class-Incremental Learning by Knowledge Distillation With Adaptive Feature Consolidation (CVPR 2022 · similarity 0.88)
- End-to-End Incremental Learning (ECCV 2018 · similarity 0.88)
- Incremental Learning via Rate Reduction (CVPR 2021 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
